Application Of Joint Time-Frequency Analysis Method In Parameter Identification

Time-frequency joint analysis method, which developed from the information engineering domain, can analyze the time-frequency characters of signal in the time and frequency plane, and is a good method for non-stationary signal analysis. It made a greatly contribution to the development of civil engineering, mechanics vibration, image processing and power system. Based on the research summarization of the time-frequency analysis methods such as linear transform, bi-linear transform and Hilbert-Huang transform in the civil engineering, this thesis proposed the time-frequency joint analysis method to identify the modal parameters of the linear, non-linear system and damage detection of structure.A method based on adaptive chirplet-based signal approximation and adaptive spectrogram is developed to identify the modal parameter of linear system in this thesis, and the signal is adaptively decomposed by the chirplet function on the sequence of the time-frequency energy. The adaptive spectrogram is obtained according to decomposition terms. The modal vibration components of response signal are firstly decomposed because thy concentrated on the vicinity of the modal frequencies, and then the noise signal is decomposed due to its comparatively scattered energy. When the most energy of modal vibration components is decomposed, then the decomposition process is ceased to avoid the influence of noise on the adaptive spectrogram. The modal parameters of system can be identified from adaptive spectrogram and the decomposition terms, and the proposed method is robust to noise.Based on the methods of empirical mode decomposition (EMD) and bi-linear time-frequency distribution, a new method to identify to modal parameters is proposed. Modal frequency and damping are identified by modal response components which separated by the method of EMD. Mode is identified by cross bi-linear time-frequency distribution of the modal response components and response signals which made it unnecessary to extract the modal response components from every measured response.The identification of non-linear system is important to obtain the characters of non-linear system, and the frequency and damping characters of non-linear system are varying with amplitude of vibration. A method of combing wavelet transform with reassignment method to extract the time-varying characters of non-linear system, and obtain frequency skeleton curve and damping skeleton curve is developed. The energy ridges of time-frequency spectrum of the signal represent the time-varying characters of the system, so accurately obtaining the ridges is important to analyze the system. The common method to calculate the ridges is based on the maximum value and optimization method, but these methods are easily affected by noise. However, the reassignment is rearranged the value the center of gravity of time-frequency and obtained the reassignment spectrogram, which has higher concentration than original time-frequency. Wavelet transform is applied to slow-varying response signal of...